基于时间序列和神经网络的主轴热误差建模

Modeling for Thermal Error Based on Time Series-Neural Network

为了减小主轴热误差影响,提高机床的加工精度,提出了基于时间序列和神经网络的主轴热误差综合模型。选用VM850立式加工中心为试验对象,对主轴的关键点温度和热误差数据进行采集,并将温度数据作为神经网络的输入变量,历史热误差和温度数据作为时间序列的输入变量,通过调整权重系数建立主轴热误差的综合预测模型。将综合预测模型和传统的时间序列模型以及神经网络模型进行对比试验,结果表明:该综合预测模型减少了关键温度点的数量,并且获得了较高的建模精度和模型鲁棒性。

In order to reduce the impact of spindle thermal error and improve the precise of the machine tool, a novel method combining times series and neural network is proposed for thermal error modeling in machine tools. Temperature and error datum are collected by a test designed on the VM850 machining cen-ter , and the thermal error compensation model is established by using temperatures as input datum in NN and using errors as input datum in TS. The results show that this novel error compensation model performs much better than the traditional TS and NN models. It requires less temperature sensors and has a good calculate precision and robustness.